Advanced machine learning for optimal parameter prediction in friction stir processing of Al-6061 alloy with alumina nanoparticle reinforcement

This study investigates the effects of friction stir processing (FSP) on Al-6061 aluminium alloy, reinforced with aluminium oxide nanoparticles. Using a CNC milling machine, various processing factors such as feed rate, number of passes, and rotational speed were explored to understand their influence on the ultimate and yield strengths, natural frequencies, and damping ratios of the samples. The processed data properties were predicted using a sophisticated machine learning technique called SRS-optimized long short-term memory (LSTME). Friction stir processing significantly enhances damping characteristics by refining the grain structure. Increasing rotational speed and traverse speed improve damping properties and mechanical characteristics. The addition of alumina nanoparticles further enhances the dampening capabilities of the material. The highest level of damping efficiency was seen when the rotational speed was set at 900 rpm for all measured passes. With an increasing number of passes as an FSP parameter, there is a decrease in the shear modulus and natural frequency, while the loss factor and damping ratio experience an increase. Higher rotational speeds generate additional thermal energy, resulting in stronger materials due to grain breakdown and increased resistance to deformation. This finer grain structure, resulting from higher rotational speeds, leads to stronger materials and higher yield strength. The developed machine learning model achieved impressive R2 values: 0.911 for ultimate strength (UTS), 0.951 for yield strength (YS), 0.953 for natural frequency, and 0.985 for damping ratio. Higher rotational speeds result in stronger materials, with a significant increase in yield strength attributed to finer grain structures.